Correcting for defects in a digital image taken by an image sensor caused by pre-existing defects in two pixels in adjacent columns of an image sensor

Abstract

A method for correcting for defects in a digital image taken by an image sensor when there are pre-existing defects in two pixels in adjacent columns of the image sensor which causes two adjacent lines of pixels in the digital image to have corrupted data.

What is claimed is:  
     
       1. A method for correcting for defects in a digital image taken by an image sensor when there are pre-existing defects in two pixels in adjacent columns of the image sensor which causes two adjacent lines of pixels in the digital image to have corrupted data, comprising the steps of: 
       (a) providing a defect map which identifies the position of the defective pixels and specifies the two adjacent lines of pixels which during readout will be caused to have corrupted data;  
       (b) capturing the digital image in the image sensor and reading out such digital image to provide the digital image with the two adjacent lines of pixels in the digital image having corrupted data;  
       (c) computing classifiers based on adjacent non-corrupted pixel data which indicate that there is a horizontal edge or a diagonal edge feature which passes through the defective lines of pixels; and  
       (d) adaptively replacing the data in the corrupted image pixels by selecting an algorithm which correponds to the edge feature identified by the classifier and using the valid data in the neighboring non-corrupted pixels of the selected edge feature.  
     
     
       2. A method for correcting for defects in a digital image taken by a full frame CCD when there is a pre-existing defect in at least one pixel in the full frame CCD which causes two adjacent lines of pixels in the digital image to have corrupted data, comprising the steps of: 
       (a) providing a defect map which identifies the position of the defective pixel in the full frame CCD and specifies the two adjacent lines of pixels which during readout will be caused to have corrupted data;  
       (b) capturing the digital image in the full frame CCD and reading out such digital image to provide the digital image with the two adjacent lines of pixels having corrupted data;  
       (c) computing classifiers based on adjacent non-corrupted pixel data which indicate that there is a horizontal edge or a diagonal edge feature which passes through the defective lines of pixels; and  
       (d) adaptively replacing the data in the corrupted image pixels by selecting an algorithm which correponds to the edge feature identified by the classifier and using the valid data in the neighboring non-corrupted pixels of the selected edge feature.  
     
     
       3. A method for correcting for defects in a digital image taken by a full frame color CCD having a plurality of different colored pixels wherein there is a pre-existing defect in at least two color pixels in the full frame color CCD which causes two adjacent lines of pixels of different colors in the digital image to have corrupted data, comprising the steps of: 
       (a) providing a defect map which identifies the position of the defective pixel in the full frame color CCD and specifies the two adjacent lines of pixels which during readout will be caused to have corrupted data;  
       (b) capturing the digital image in the full frame color CCD and reading out such digital image to provide the colored digital image with the two adjacent lines of pixels having corrupted data;  
       (c) computing classifiers based on adjacent non-corrupted pixel data which indicate that there is a horizontal edge or a diagonal edge feature which passes through the defective lines of pixels; and  
       (d) adaptively replacing the data in the corrupted image pixels by selecting an algorithm which correponds to the edge feature identified by the classifier and using the valid data in the neighboring non-corrupted pixels of the selected edge feature.  
     
     
       4. The method of  claim 3  wherein the full frame color CCD is provided in a single sensor digital camera which produces single color values for each color pixel in the digital image and includes the step of: 
       (e) interpolating color pixels in the digital image to provide for a plurality of color values for each color pixel in the digital image and wherein step (d) is provided before or after the interpolation step.  
     
     
       5. The method of  claim 3  wherein the replacing step includes providing a plurality of classifiers based upon neighboring non-corrupted color pixels and selecting at least one of the classifiers to identify an appropriate formula for operating on neighboring non-corrupted color pixels to replace the values of the corrupted color pixels in the captured image. 
     
     
       6. A method for correcting for defects in a digital image taken by a full frame color CCD having a plurality of different colored pixels wherein there is a pre-existing defect in at least two color pixels in the full frame color CCD which causes two adjacent lines of pixels of different colors in the digital image to have adjacent corrupted data that are two pixels wide and at least a portion of one of the lines to have corrupted data which are one pixel wide, comprising the steps of: 
       (a) providing a defect map which identifies the position of the defective pixel in the full frame color CCD and specifies the two adjacent lines of pixels which during readout will be caused to have corrupted data;  
       (b) capturing the digital image in the full frame color CCD and reading out such digital image to provide the colored digital image with the two adjacent lines of pixels having corrupted data;  
       (c) computing classifiers based on adjacent non-corrupted pixel data which indicate that there is a horizontal edge or a diagonal edge feature which passes through the defective lines of pixels for the two pixels wide corrupted data or the one pixel wide corrupted data; and  
       (d) adaptively replacing the data in the corrupted image pixels by selecting an algorithm which correponds to the edge feature identified by the classifier and using the valid data in the neighboring non-corrupted pixels of the selected edge feature.